The influence of mineral admixtures on the rheology of cement paste and concrete

The influence of mineral admixtures on the rheology of cementpaste and concreteChiara F.Ferraris a,*,Karthik H.Obla b,Russell Hill ba NIST,100Bureau Drive,Mail Stop8621,Gaithersburg,MD20899,USAb Boral Material Technologies,45N.E.Loop410,Suite700,San Antonio,TX78216,USAReceived3March2000;accepted12October2000AbstractRheological tests on cement paste were used to successfully select the type and dosage of mineral admixtures that improved concrete workability.Among the six different mineral admixtures tested,the ultrafine fly ash(UFFA)was determined to give the best results by reducing the yield stress and viscosity.These improved rheological properties were not achieved by increasing the water demand and/or the high-range water reducer(HRWR)admixtures dosage.Therefore,the addition of UFFA improved the concrete flow without a potential decrease of the hardened properties or an increase in cost.The conclusions reached based on cement paste tests were validated by concrete slump tests.The cement paste rheological data were also compared using two simpler tests,minislump and Marsh cone.The goal was to determine whether the simpler tests could be used to characterize the rheology of cement paste adequately.The conclusions are that these simpler tests are unreliable for measuring workability.D2001Elsevier Science Ltd.All rights reserved.Keywords:Rheology;Cement paste;Workability;Fresh concrete1.IntroductionHigh-performance concrete(HPC)is a complex mixture often containing5±10different materials.Interaction between the various materials can cause wide variations in workability,which also depends on the specific materials and proportions used.Determining workability properties by testing concrete is not always practical.Extensive con-crete testing requires a large amount of materials and labor, which is expensive.There is,therefore,a need to predict the workability of concrete through a simpler,cheaper labora-tory approach.This paper describes an attempt to utilize rheological measurements in cement paste as a reasonable indicator of concrete workability.Concrete workability is defined,according to American Concrete Institute(ACI),as the ease of placement of concrete and is usually quantified by the result of the slump cone test.Rheological tests on cement paste were used to select the type and dosage of mineral admixtures that improved concrete workability.The conclusions reached based on cement paste tests were validated by concrete slump tests. It was also deemed interesting to compare the fundamental parameters(yield stress and viscosity)measured with a fluid rheometer with the results from two commonly used empiri-cal tests,the minislump and the Marsh cone tests.If a relationship could be established,the empirical tests could be used to design materials for a given yield stress and viscosity or,at the very least,rank different materials base on yield stress or viscosity.The decision to study the influence of mineral admixtures was dictated by the recent increase in use of mineral admixtures for improved concrete durability.Economics (lower cement requirement)and environmental considera-tions have also had a role in the growth of mineral admixture usage.The lower cement requirement also leads to a reduction in the amount of carbon dioxide generated by the production of cement,while the use of a mineral admixture utilizes a product that would ordinarily be bound for the land fill.Thus,there is a double environmental benefit from using mineral admixtures.*Corresponding author.Tel.:+1-301-975-6711;fax:+1-301-990-6891.E-mail address:clarissa@(C.F.Ferraris).Cement and Concrete Research31(2001)245±2550008-8846/01/$±see front matter D2001Elsevier Science Ltd.All rights reserved.PII:S0008-8846(00)00454-32.Background2.1.Concrete workability characterizationConcrete workability cannot necessarily be sacrificed for improved hardened properties,such as durability or strength.Workability is typically quantified in the field by the result of the slump cone test.Nevertheless,a survey[1] conducted by the National Ready-Mixed Concrete Association(NRMCA)and the National Institute of Standards and Technology(NIST)determined that,for HPC,the slump cone value is not representative of the ease of handling HPC in the field.It was reported that concrete mixtures with the same slump might not behave the same way during placement.This implies that the slump value does not give enough data to fully characte-rize concrete flow.In the construction field,terms like workability,flow-ability,and cohesion are used,sometimes interchangeably, to describe the behavior of concrete under flow.The definitions of these terms are very subjective.Therefore, there is a need for a more fundamental and quantitative description of concrete flow.Rheological measurements of concentrated suspensions can be used to describe the flow of concrete.Numerous researchers[2±4]have successfully used the Bingham equation.Two parameters define the flow:yield stress and plastic viscosity[5].Yield stress is related to slump[6,7],but plastic viscosity is usually ignored because only a few type of instruments exist to measure it[8].However,the viscosity may be related to properties such as stickiness,placeability,pumpability,and finishability.In addition,segregation could be defined as the ability of the aggregate to migrate(or sink)in the cement paste.This phenomenon is linked to the viscosity of the cement paste and the concrete mixture design.Therefore, methods to predict concrete workability need to take into account more than just the yield stress.2.2.Concrete rheology from cement paste rheologyAdmixtures mainly affect the flow behavior of the cement paste without altering the composition or behavior of the aggregates.Therefore,it seems reasonable to try to select admixtures,chemical and mineral,by only testing the cement paste.Ideally,the results will then be related to the concrete workability.Unfortunately,the relationship between cement paste rheology and concrete rheology has never been completely established.The main reason for this is that cement paste rheology is typically measured under conditions that are never experienced by the cement paste in concrete.Therefore,the measured cement paste rheological parameters may differ from those estimated from concrete parameters.The values usually reported in the literature for cement paste do not take into account the contribution of the aggregates[9].The aggregates act as heat sinks and shear the cement paste during the mixing process.A computerized model to simulate the concrete being sheared is under development at NIST[10].This model will predict concrete rheology from constituent properties that will include rheological measurements on cement paste.However,the cement paste flow properties,if measured ``properly,''can be used to screen mineral admixtures.The details of the methodology are presented elsewhere[11], but the following principles should be reiterated.The cement paste needs to be mixed and tested under the conditions similar to those that it will experience in concrete,primarily shear and temperature history.There-fore,a high-shear,temperature-controlled mixer should be used.To measure the rheological parameters of cement paste flow,a parallel plate rheometer must be used because it is the only rheometer with variable geometry.It has been established[12]that the rheological properties of cement paste change if the material is squeezed between two surfaces or aggregates as in concrete.The distance between the surfaces is called the``gap.''In a parallel plate fluid rheometer,the gap or distance between the plates can be varied easily to more closely represent the shearing action imposed on paste in concrete.In this paper,cement paste rheological parameters were measured using the above method as a function of type and dosage of mineral admixture.2.3.Role of fine powder additions on workabilityIt is usually reported that,if the volume concentration of a solid is held constant,the addition of mineral admixtures improves concrete performance but reduces workability.The most common reason for poor work-ability is that the addition of a fine powder will increase the water demand due to the increase in surface area.This belief is supported by test results that show that the addition of fine silica fume(SF)particles increases the water demand to attain specific workability levels.How-ever,in certain cases,it is reported in the literature that the use of fine mineral admixtures can reduce the water demand or increase the nge et al.[13]measured the water demand of mortars with increasing additions of a very fine blast furnace slag.He found that,for a specific flow,an optimum amount of blast furnace slag reduced the water demand of the mortar.A popular hypothesis put forward to explain the workability enhancement due to the use of certain fine mineral admixtures,especially fly ash (FA)or SF,is that the spherical particles easily roll over one another,reducing interparticle friction[14].The sphe-rical shape also minimizes the particle's surface to volume ratio,resulting in low fluid demands.Out of all3D shapes, a sphere gives the minimum surface area for a given volume[15].Sakai et al.[16]reported that a higher packing density was obtained with spherical particles as compared to crushed particles in a wet state.This resulted in lower water retention in the spherical case and subse-C.F.Ferraris et al./Cement and Concrete Research31(2001)245±255 246quently lower water demand for a specific workability.A strong dependence of fluidity(defined as the inverse value of the viscosity)on the average particle size was reported with a pessimum value[16].It was explained that,at an optimal particle size,the packing density was maximum, which helped to achieve maximum fluidity.Recently, Collins and Sanjayan[17]reported that in concrete con-taining alkali-activated ground granulated slag as the binder,the workability was improved by replacing partof the binder with ultrafine materials.This material had 90%by mass of the particles smaller than13.7m m.It was also reported that some similar materials were not effective in improving the workability.It can be concluded from this survey of the literature that the selection of a fine mineral admixture for improved concrete workability is not a trivial problem.At present,this selection cannot be predicted from the physical or chemical characteristics of the admixture,and can only be determined using a properly designed test.3.Materials,mixing,and testing detailsThe materials,mixing,and testing requirements provided here are for the cement paste tests only.The details for the concrete test program are provided in Section5.3.1.MaterialsThe cement was an ASTM Type I portland cement whose composition is described in Table1.This cement was used for all tests,both concrete and cement paste.The high-range water reducer(HRWR)was a naphthalene sulfonate-based product with a mass fraction of43%active ingredients.The mineral admixtures used are shown on Table2with their mean particle diameters(PDs).The mean PD was measured using a laser diffraction particle size analyzer.Four different FAs,all from the same plant, were tested.FA was the standard FA available in that plant and used in concrete type applications.Coarse fly ash (CFA)was a coarse ash that is usually rejected as it does not meet the ASTM C618requirements for particle size. Fine fly ash(FFA)is a finer form of ash obtained by separation from a coarser ash using a classifier.Ultrafine fly ash(UFFA)is an ultrafine ash obtained by still more rigorous separation.The cement paste composition was varied to explore the influence of mineral admixture dosage and type on the rheological properties.The performance differences in the paste due to the addition of mineral admixture were mea-sured either by the rheological properties at constant water content or by the water reduction at constant mineral and chemical admixture dosage.The compositions of the cement pastes can be summarized as follows:water/cement ratio:0.28±0.35;dosages of mineral admixtures:0±16%of cement,replacing cement by mass;anddosage of HRWR(based on naphthalene sulfonatedcondensate):0.45±0.70%solid by mass of cementi-tious materials.3.2.Cement paste preparationThe cement paste preparation is very important because the shear history of a mixture will influence its rheological behavior.In this case,we wanted to have the same shear history as in the concrete in order to be able to compare the neat cement paste behavior and the concrete flow.Two types of mixers were considered:a standard paddle mixer[18]and a high-speed blender.The cement paste was mixed using a paddle mixer according to the procedure of ASTM C305, except that no sand was added.The second mixer used was a large(4L)blender.This blender was not temperature controlled.The cement paste mixture was prepared according to the following procedure.The cement and the mineral admixture(if any)weremixed dry for about5s by hand outside the mixer. Water was poured into the4-L blender.The mixer was started at slow speed and the cement±mineral admixture mixture was added over a50-speriod.The HRWR was added in5s.The mixture was mixed for another60s at slow speed.Table1Cement compositionChemical composition Percentage by mass Loss on ignition(LOI) 1.29Sulfur trioxide(SO3) 2.79Silica dioxide(SiO2)20.86Ferric oxide(Fe2O3) 3.47 Magnesium oxide(MgO) 1.21 Aluminum oxide(Al2O3) 4.60 Equivalent alkalies(as Na2O)0.46Calcium oxide(CaO)64.34Free lime±Insoluble residue(IR)0.1 Tricalcium silicate(C3S)59.57 Tricalcium aluminate(C3A) 6.31Table2Mineral admixtures Mean PD[m m] 1Coarse Fly Ash(CFA)18.02Fly Ash(FA)10.96Fine Fly Ash(FFA) 5.73Ultra Fine Fly Ash(UFFA) 3.14Metakaolin(MK)7.45Silica Fume(SF)%0.1C.F.Ferraris et al./Cement and Concrete Research31(2001)245±2552473.3.Cement paste testing detailsFresh cement paste specimens were tested first in a parallel plate fluid rheometer and then using the minislump and Marsh cone.The testing details are given below.3.3.1.Parallel plate rheometerThe parallel plate rheometer was used to determine the yield stress and the plastic viscosity as defined by Bingham.The distance between the two plates of the parallel plate rheometer should be selected based on the cement paste content in the concrete mixture to be characterized.How-ever,as this method was used here as a screening test,and no comparison was sought with a specific concrete mixture design,the gap (distance between the plates)was arbitrarily fixed at 0.4mm,based on the median value of distance between aggregates in concrete [11].The shear rate used ranged from 3to 50s À1.This range was selected to correspond to the shear rates used in a concrete rheometer [7].The surfaces of the two plates were serrated,as provided by the manufacturer,to avoid slippage.The following is a list of the sequence of the measurements done..One milliliter of cement paste was placed,using a syringe,on the bottom plate..The two plates were brought together to the desired distance of 0.4mm..The computer driven system imposed a slowly increas-ing shear rate from 0to 70s À1in 160s.As soon as the highest shear rate was reached,the plate stopped rotating..After this first phase (needed to homogenize the specimen),a full cycle of increasing shear rate by 10steps from 3to 50s À1and back to 0shear rate with another 10steps was performed.At each step,the stress was measured and the value of constant stress was recorded.If the constant stress was not achieved in 20s,the computer took the average of the last five values recorded..The slope of the down-curve (decreasing shear rate)was used to calculate the plastic viscosity,while the inter-cept at zero shear rate was used to calculate the yield stress.An example of the curve obtained is shown in Fig.1.3.3.2.Minislump cone test detailsKantro [19]developed the minislump cone test.The test was conducted as follows:A square piece of flat glass on which the diagonals and the median were traced was covered with a plastic sheet.The plastic sheet is needed to avoid water or very fluid paste leakage under the minicone.The minislump cone was placed in the center of the glass plate and filled with the cement paste.The cone was gently lifted and after one minute the diameter of the pad formed was measured along the marked line on the glass in four directions.All four diameters were recorded and the average diameter was calculated.3.3.3.Marsh cone test detailsThe Marsh cone is a funnel with a long neck and an opening of 5mm,currently used to test oil well cement [20].This test is not a standard test.The test was conducted using the following procedure..A Marsh cone was attached to a stand so that the small orifice was pointing down and a glass graduated cylinder was placed under the cone..Closing the small orifice with a finger,1L of cement paste was poured into the cone..The orifice was opened and a stop watch started..The time for a certain amount of cement paste to flow was recorded.The volumes selected were 300,500,and 700mL.These settings were selected from research conducted by Nehdi et al.[21]that showed a nonlinear flow of the cement paste for amounts higher than 700mL.4.Results and discussion4.1.Effect of mixer type on rheological properties of cement pasteThe goal of testing cement paste instead of concrete is to save materials and labor.However,to be useful,the cement paste results need to predict concrete performance.There-fore,the cement paste needs to be sheared with about the same intensity,as it would have experienced while being mixed in concrete.One way to determine if the mixing method selected is appropriate is to select several cement paste compositions and compare the rheological behavior of the cement paste with the concrete performance.In this paper,we selected cement paste with and without mineral admixtures.The key result,which would tell us that we had the correct mixing method,is that the mineral admixture that resulted in the larger reduction in yield stress or viscosity compared to the control,which had no mineral admixture,would be the same in cement paste and in concrete.The two mixers available to us were a Hobart and ablender.Fig.1.Typical curve obtained from the fluid rheometer.C.F .Ferraris et al./Cement and Concrete Research 31(2001)245±255248To perform the comparison,we selected three mineral admixtures(metakaolin(MK),UFFA,and SF)to test in cement pastes.If the selection of the best mineral admixture was based on the data shown in Fig.2,the result would depend on the mixer used.Consider the yield stress(YS) behavior(bars and left axis).The best admixture should reduce the yield stress compared to the control(no admix-tures).If the black bars(paddle mixer)are examined,the best admixture will be the MK,while if the gray bars are examined,the choice will be UFFA.In examining the results obtained with concrete[23],for equivalent slumps, the water demand is the lowest for mixtures containing UFFA.In other words,UFFA increases slump in concrete when water content is kept constant.In conclusion,UFFA is the``best''admixture both in concrete and in cement paste if mixed in a blender.Therefore,concrete performance was more accurately predicted by the cement paste mixed in the blender and not in the Hobart mixer.This confirms a studyby Helmuth et al.[22]stating that in concrete,during mixing,the cement paste is sheared with an energy and rate more closely reproduced in a blender as opposed to the low shear rate of the Hobart mixer.Therefore,to predict concrete behavior,it is essential to use the correct mixer while preparing cement paste.Thus,the rest of the data reported in this paper were obtained using the blender.parisons between rheological and empirical testsA fluid rheometer for cement paste is not widely avail-able in the construction industry for many reasons.The two main reasons are:(1)the instrument is relatively expensive (on the order of US$40,000)and(2)the importance of using such a device for cement paste was not advocated until recently[11,12].Therefore,it would be advantageous to be able to use simpler tests such as the minislump and the Marsh cone tests.A comparison between the rheometer and the other test results is presented in Figs.3and4.These figures are a compilation of all the tests conducted in this research program using a blender to mix the cement paste. Each point represents various mineral admixture additions at various dosages,W/C,and HRWR dosages.The plot of yield stress vs.minislump spread diameter (Fig.3)shows a weak correlation:Higher yield stress corresponds to a lower spread in the minislump.There-fore,an indication of the yield stress could be obtained using the minislump.This result was expected because the cement paste in a minislump test will only flow if the stress due to the weight of the cement paste contained in the cone is high enough,i.e.,higher than the yield stress of the cement paste.It should be kept in mind that the minimum diameter that can be measured is70mm, corresponding to the diameter of the bottom of the cone. Therefore,despite some of the scatter of the data shown on Fig.3,an approximation of the yield stress could be obtained by fitting a straight line through the data.This parison between the minislump spread and yield stress.Each point represents a measurement(not an average),therefore,no estimate of the uncertainty can becalculated.Fig.2.Influence of the mixer on the rheological properties of cement paste.``YS''is the yield stress and``Visc.''is the plasticviscosity.parison between the marsh flow time and plastic viscosity.Eachpoint represents a measurement(not an average),therefore,no estimate ofthe uncertainty can be calculated.C.F.Ferraris et al./Cement and Concrete Research31(2001)245±255249fit will not be done here because it has a limited significance due to the wide scatter of the data.In contrast,the plot of time to flow for300ml(Marsh cone flow test result)and the viscosity(Fig.4)shows no correlation at all.If outliers are removed,a``shot gun'' distribution can be observed.Similar results were obtained when the times to flow for500or700ml were plotted, because the time to flow had a linear relationship with the amount of material measured.Nevertheless,in some limited cases,a lower time to flow does correspond to a lower viscosity.However,it would be dangerous to rely on the Marsh cone to select a material for a certain viscosity requirement or even to rank materials based on viscosity due to the overall lack of correlation.This result is some-what unexpected because it was assumed that the weight of the cement paste was high enough to overcome the yield stress and therefore the speed of the cement paste flow through the flow cone would depend on its viscosity.From the results obtained,it seems that other factors contribute to the flow,such as friction and sedimentation.4.3.Effect of mineral admixture type on cement paste rheological propertiesIn Fig.5,the yield stress and viscosity are shown for mixtures composed of cement paste with the same W/C ratio of0.35and varying dosages of HRWR.The amounts of the various mineral admixtures by mass as replacement of cement are indicated on the figure.It is clear that the replacement of cement with UFFA leads to a decrease in the HRWR dosage over the control(no mineral admixtures)at a given yield stress or viscosity.In contrast,the replacement of cement by SF significantly increases the HRWR dosage at a given yield stress and viscosity.The addition of MK shows no significant improvement in yield stress and plastic viscosity over the control.Therefore,there are no significant rheological benefits or drawbacks in using MK as a mineral admix-ture,at least at the dosages tested.4.4.Effect of PD on cement paste rheological properties In Fig.6,the rheological measurements for the four FA/cement pastes are plotted against the mean particle size of the FAs.All tests were conducted at the same dosage of mineral admixture(12%replacement of cement by mass),same W/C ratio(0.35)and same dosage of HRWR(0.45%solid by mass of cement).It is clear that the lowest yield stress and viscosity are obtained at a mean PD of3m m.This value corresponds again to UFFA.It also seems that maximum viscosity is reached at a mean PD of about11m m,and maximum yield stress at a mean PD of5.7m m.This result seems to indicate an optimum and a pessimum PD,with the optimum at3m m and the pessimum at5.7m m.Unfortunately,a FA with a smaller PD than3m m was not available to determine if the correct optimum was reached.Sakai et al.[16]also showed that there is a pessimum at18m m,but he used limestone powder and not FA.It is conceivable that the optimum and pessimum value depends on the type of mineral admixture used,and the chemistry and physics of the individual particles.4.5.Effect of W/C on rheological properties of cement paste Fig.7shows the results of tests performed on cement pastes with UFFA(at12%replacement)at various W/C ratios,plotted vs.HRWR dosage.There are several ways to use or to interpret these results:(1)determine the correct dosage of HRWR needed to obtain the same yield stress and/or viscosity with the UFFA mixes and the control at various W/C;(2)determine the water reduction achieved by using UFFA and maintaining the same yield stress and/or viscosity;or(3)determine the reduction inHRWRFig.5.Dosage of HRWR and its effect effect on the flow of properties.The W/C ratio was0.35.The error bars represent an estimate percentage error:1.7%on yield stress and10%viscosity.This error was estimated from the numerous tests done.C.F.Ferraris et al./Cement and Concrete Research31(2001)245±255250dosage achieved while maintaining the same yield stress and/or viscosity.In summary,the addition of UFFA improves the rheolo-gical properties.If the goal is to add UFFA and achieve the same yield stress and viscosity as the control,Fig.7shows that the W/C ratio can be reduced by 10%and the HRWR dosage can be reduced by 40%.On the other hand,if the water content is reduced by 20%(W/C ratio of 0.28)a significant increase of the HRWR dosage (almost double)is needed to maintain the yield stress or viscosity,giving the same rheological behavior as the SF mixes.4.6.Effect of dosage of UFF A on cement paste rheological propertiesFig.8shows the influence of dosage of UFFA on cement paste rheological properties.The tests were done at a W/C ratio of 0.35and a fixed HRWR dosage of 0.45%solid by mass of cement.The plot suggests that a dosage of 12%bymass is optimal for the best rheological properties.The dosage shows an optimum value corresponding to the lowest value achieved by the yield stress for a 12%UFFA by mass dosage.The shape of the curve (Fig.8)is an important result because it corresponds to the same type of behavior seen in concrete,as will be shown in Section 6and in Fig.10.5.Concrete testing 5.1.Concrete mixtureThe purpose of the concrete measurements was to validate the conclusions drawn from the data obtained in the cement paste for the best type and dosage of mineral admixture.The concrete rheological behavior was tested using the standard slump cone test.Therefore,only an indication of the yield stress was available forcomparison.Fig.6.Influence of mean PD on the flow properties of cement paste.The W/C ratio was 0.35.The error bars represent an estimate percentage error:17%on yield stress and 10%viscosity.This error was estimated from the numerous testsdone.Fig.7.Influence of W/C ratio on the rheological properties of cement paste with UFFA at 12%replacement of cement by mass.The numbers in the legend indicate the W/C ratio used.The error bars represent an estimate percentage error:17%on yield stress and 10%viscosity.This error was estimated from the numerous testsdone.Fig.8.Influence of the dosage of UFFA on the rheological properties of cement paste with constant dosage of HRWRA (0.44%(13oz/100lb of cementitious materials)).The W/C ratio was 0.35.The error bars represent an estimate percentage error:17%on yield stress and 10%viscosity.This error was estimated from the numerous tests done.C.F .Ferraris et al./Cement and Concrete Research 31(2001)245±255251。